3x GeForce4 Ti from Inno3D in the test: Cheaper range of models

3x GeForce4 Ti from Inno3D in the test: Cheaper range of models


Size comparison

We don't want to go any further here on the technology , since this has already been discussed in great detail, including in our first GeForce4 Review the Asus cards and in the GeForce4 Preview . Just a little about the size. As is known, the GeForce4 cards are anything but small. On a board from Epox there was even a 'hardware conflict' between the GeForce4 4600 and a capacitor that is attached directly to the AGP port. As you can see in the photo above, the GeForce4 Ti 4200 (middle) is significantly smaller in size than the Ti 4400 (above) and also the Ti 4600 of the same size. Both, however, look almost dwarfed next to the huge PCB (printed circuit board) of the Voodoo5 5500 by 3dfx (below).

Test system

This time a system based test came up with the test from a combination of SiS745 chipset and AthlonXP 2100+ with 256MB DDR-RAM are used. We have thus taken into account the necessary upgrade promised last time and removed a little more limitation for the performance of the graphics card.

We tested under Windows2000, as it is the basis for the home OS of the future in the portfolio from Microsoft and therefore it is easy to estimate how reliable and high-performance the respective drivers behave under this operating system.

Service Pack 2 and the current application update from Microsoft were installed as well as chipset drivers of the version 1.09 from SiS. The latest DirectX 8.1 goes without saying. The values ​​for the GeForce3 and GeForce4 Ti cards were determined with the almost current driver 29.40 from nVidia, which shows almost no changes to the current reference driver 29.42.

For background noise For a change, the subjectively satisfactory onboard sound of the SiS chipset was used with the 7012.1.03c drivers.

We are aware that with a faster base system and without the soundscape, the differences between the GeForce3 and GeForce4 Ti as well as between the individual GeForce4 Ti modelscould. So what is shown here is not to be understood as an ideal scenario. The overall performance and in particular the performance of the GeForce4 Ti could therefore increase significantly.

  • CPU:
    • AMD AthlonXP 2100+ (1733MHz)
  • Motherboard:
    • Elitegroup K7S6A with SiS745 chipset
  • RAM:
    • 1 * 256MB KingMax BGA PC2700 DDR -RAM CL2,0
  • Graphics card:
    • Inno3D Tornado GeForce 4 Ti 4400
    • Inno3D Tornado GeForce4 Ti 4200 (64MB)
    • Inno3D Tornado GeForce4 Ti 4200 (128MB)
    • GeForce3
  • Peripherals
    • IBM DTLA 307015
    • Pioneer DVD A03
    • Realtek RTL8139c network card (deactivated)
    • Castlewood ORB removable drive
  • Other
    • A4 Tech Double Wheel Mouse
  • Software
    • Windows2000 Professional
    • Service Pack 2
    • Application Update
    • DirectX 8.1


We once again opted for a somewhat broader range of tests en. On the one hand, to get a more representative overview of the performance spectrum of the cards and, on the other hand, to counteract the special driver optimizations of the chip manufacturers for the two most popular benchmarks, Quake3 and 3DMark2001, which have recently become more and more popular. In this high-price segment for enthusiasts in particular, it would be very disappointing if real improvements were only noticeable in the two or three standard benchmarks.

Synthetic benchmarks
  • 3DMark 2000 v1.1
  • 3DMark 2001 SE
  • Templemark D3D v1.06
  • Villagemark D3D v1.19
  • GL Excess v1.1a
  • VulpineGL v1.1p
  • Codecreatures Bench
  • Aquamark
Game benchmarks
  • Comanche 4
  • DronezMark
  • Aquanox v1.17
  • Dungeon Siege
  • Giantsv1.4
  • Unreal Tournament v4.36
  • Quake III Arena v1.30
  • Jedi Knight II
  • Serious Sam v1.02
  • Max Payne v1.02
Benchmarks with FSAA and aniso. Filter
  • Quake III Arena v1.30
  • 3DMark 2001 SE
  • Dungeon Siege

Unless otherwise noted, all benchmarks ran in 32-bit color depth and, if possible, also 32-bit textures. In general, the highest possible detail settings were used to really challenge the test subjects.

On the next page: Synthetic benchmarks